Analogue multiplexer diagram.

agerasiaetherealAI and Robotics

Nov 24, 2013 (3 years and 8 months ago)

87 views


Impact of AC
-
coupling on the SSD performance.


Report given at ITS meeting at CERN on 5.12.2001.


Vladimir Gromov

(vgromov@nikhef.nl)

Department of electronics,

NIKHEF, Amsterdam, the Netherlands.


. due to use of double
-
sided structure, the front
-
end electronics of SSD on both
sides operate at different potentials (detector bias


55


⸠瑨攠晲o湴
-
敮搠捨ipso映瑨攠摥瑥捴omo摵敳a攠牥ao畴ui渠s敲iesⰠbo瑨t瑨t
-

a湤nN⁳楤攠o映fa捨⁤整散eo爠o摵攠eo攠eDC⁣桡湮敬with‱M䡚⁲慴⸠


⸠AC
-
捯異
li湧is瑨攠o湬ywayto灲pvi摥瑨攠signal瑲ansmissio渮n䡯weve爠it
牥ul瑳i渠nig湡l⁤isto牴io渠inasm畣栠慳⁴桥⁳楧halg整s⁤楦晥牥湴na瑥搠t捣o牤湧⁴o

F⡰⤽(

/⠱⭰

⤬w桥牥

㵃=1⸲k



Analogue multiplexer diagram.


P
-
side

P
-
side

N
-
side


N
-
side

to ADC

. the coupling capacitor value should be kept as low as possible for
safety and
space reasons. On the other hand the smaller the capacitor is the greater signal
distortion becomes. That leads to the detector performance deterioration.

What is the problem with AC
-
coupling?

AC
-
coupling in front of analog buffer changes the sh
ape of the signal coming out of the front
-
end chip
(HAL25).


Distortion of the hit charge could leads to wrong position information,



Capacitor is bigger.



AC
-
coupling causes extra ineffi
ciency.


mistake

Strip#N

Str
ip#N+1

Strip#N+5

Strip#N+6

Signal before
AC
-
coupling

Exponential tail

mistake

mistake

mistake

Exp(
-
t/t
d
)

Signal after AC
-
coupling

Strip#N+1

Strip#N

Strip#N+5

Strip#N+6

mistake

Exp(
-
t/t
d
*
)

Threshold =2000e

The signals turn to be
below the threshold
causing inefficiency

Method is Monte
-
Karlo simulations.

Items taken into account:


1. number of channels (strips) to be read out at every TRIGGER coming in
6*128=768


2. number of hits for a TRIGGER is Poisson statistics with average of
occupancy*768



3. charge deposit to the detector by a hit is distributed according to
Landau

with most probable value

(MIP) of 22000e (41.8)



4.

threshold is set at 5*400e
(ENC)=2000e

(3.7).


5. noise contribution to a signal variation is not taken into account.


6. pitch (strip
-
to
-
strip distance) is
100um
.


7. .deposit charge is eaten by one strip if an hit interaction point is wit
hin 30um from center of the
strip
(digital zone).


8. deposit charge is shared between two neighbouring strips if an hit interaction point is within the
area from 30um to 70um from center of a strip
( analog zone).















9. AC coupling is s
imulated with response of differentiation circuit


where t
-

current time,


-

circuit time constant,


-

delay time,

(t)=1, if t>0,

(t)=0, if t<0.






10. the signal is sampled with 90ns delay in respect to the leading edge of it.



11. position of the hit is determined by
central gravity method
.



Example.









Hit#4

Position
4
=15.41

Charge
4
=51

Charge depo
sit
in strip#16

Charge
16
=13

Charge deposit
in strip#15

Charge
16
=38

Hit#3

Position
3
=10.19

Charge
3
=40

Charge deposit
in strip#10

Charge
10
=40

Hi#2

Position
2
=4.35

Charge
2
=75

Hi#1

Position
1
=4.59

Charge
1
=43

Charge deposit
in strip#5

Charge
5
=17

Charge deposit
in strip#4

Charge
4
=89

Charge de
posit
in strip#3

Charge
3
=12

Sampled value
in strip#16

S
16
=9

Sampled value
in strip#15

S
15
=35

S
ampled value
in strip#10

S
10
=37

Sampled value
in strip#5

S
5
=15

Sampled value
in strip#4

S
4
=85

Sampled value
in strip#3

S
3
=11

ERR
n

= Position

n

-



X
k
n


S
k
n
+ X
k+1
n


S
k+1
n



S
k
n
+

S
k+1
n

ERRQ
n

= Charge

n

-

[S
k
n
+

S
k+1
n
]

Input of the AC
-
circuit.

Output of the AC
-
circuit.

Results: Position resolution distortion due to AC
-
coupling.

Runs=100. Occupancy=5%.


Entries

Almost ideal case



= 1ms , C =
1000nF



occu
pancy=5%



= 1us , C =
1000pF


occupancy=5%



= 150ns , C =
150pF


occupancy=5%

Entries


m


m


m

Entries


Results:

Position resolution distortion due to AC
-
coupling.

Runs=100. Occupancy=5%, 7.5%, 10%.


Almost ideal case



= 1ms , C =
1000nF



occupancy=5%




= 1ms , C =
1000nF



occupancy=10%

Entries

Entries

Entries


m


m


m




= 1ms , C =
1000nF



occupancy=7.5%


Results: Amplitude distribution distortion due to AC
-
coupling.

Runs=100. Occupancy=5%.




Entries

Entries

Entries

Arbitrary units

Arbitrary units

Arbitrary units




= 1ms , C =
1000nF occupancy=5%



= 150ns , C =
150pF


occupancy=5%

Initial Landau
distribution



= 1us , C =
1000pF


occupancy=5%

Initial Landau
distribution

Initial Landau
distribut
ion

Results: Amplitude dis
tribution due to AC
-
coupling.

Runs=100. Occupancy=5%, 7.5%, 10%.




= 1ms , C =
1000nF occupancy=5%



= 1ms , C =
1000nF occupancy=7.5%



= 1ms , C =
1000nF occupancy=10%

Arbitrary units

Arbitrary units

Arbitrary units

Entries

Entries

Entries

Initial Landau
distribution

Initial Landau
distribution

Initial Landau
distribution

Results: Base line fluctuation due to AC
-
coupling.

Runs=100. Occupancy=5%.









Arbitrary units


=㄰ふ献⁃=㄰のF⹏ccu灡湣y=5%

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.5
=1.3

el
=520e

Base line fluctuation


bl
=346e,
mean=533e

Electronic noise


bl
=400e, mean=0

Base line fluctuation


bl
=107e,
mean=110e

Electronic noise


bl
=400e,

mean=0

Arbitrary units

Arbitrary units

Electronic noise


bl
=400e, mean=0

Base

line fluctuation


bl
=64e, mean=40e


=㔰ふ献⁃=㔰のF⹏ccu灡湣y=5%

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.5
=1.04

el
=416e


=㄰〰畳⸠1=㄰〰1F⹏cc異u湣y=㔥

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.5
=1.01

el
=
404e

Entries

Entries

Entries

Results: Base line fluctuation due to A
C
-
coupling.

Runs=100. Occupancy=5%, 7.5%, 10%.








Arbitrary units

Arbitrary units

Arbitrary units


=㔰ふ献⁃=㔰のF⹏ccu灡湣y=5%

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.5
=1.04

el
=416e


=㔰ふ献⁃=㔰のF⹏ccu灡湣y=㜮7%

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.
5
=1.08

el
=432e


=㔰ふ献⁃=㔰のF⹏ccu灡湣y=㄰%

呯瑡氠湯l獥


tot
=[

el
2
+

bl
2
]
0.5
=1.13

el
=452e

Electronic noise


bl
=400e, mean=0

Electronic noise


bl
=400e, mean=0

Electronic noise


bl
=400e, mean=0

Base line fluctuation


bl
=107e,
mean=110e

Base line fluctuation


bl
=160e,
mean=160e

Base line fluctuation


bl
=213e, mean=213e

Entries

Entries

Entries




Conclusion.






1.
Hit position information is almost not distorted

by AC
-
coupling. Position resolution remains almost the same when the
coupling capacitors are in range down to 150
pf (

=150ns) used even
under relatively high occupancies of 10%.





2.
Amplitude information has been heavily distorted

when AC
circuit with small time constant is used (below 1ms, capacitor is
1000nF). No considerable amplitude distortion can be notic
ed even
under 10% occupancy if capacitor of 1000nF is used.




3
. AC
-
coupling causes base line fluctuation and hence extra noise
.

If the coupling capacitor is 500nF (

=500us) the total noise increases
by 13% in the worse case when occupancy is 10%.


Resul
ts on AC
-
coupled ALABUF testing.

14.10.2002.

Vladimir Gromov.

NIKHEF, Amsterdam, the Netherlands.


Objectives of the testing
.

By doing measurements with a real set of signals I am going to confirm Monte
-
Karlo
simulation earlier carried out. I will prove th
at:

a). additional noise (base line fluctuations) occurs due to ac
-
coupling in front of the
ALABUF chip.

b). the smaller nominal of the coupling capacitor the bigger the additional noise is.

c) the smallest acceptable (additional noise is negligible in c
omparison with expected
electronic noise of HAL25 chip) value of the capacitor is below 100nF.


Introduction
.

According to Monte
-
Karlo simulations ac
-
coupling causes signal distortion and base line
fluctuations while fast (10MHz) analog signal read is goin
g on. The fluctuations slightly modulate the
red
-
out signal giving a deviation from the initial value. Such a deviation can be interpreted as an
additional noise and assigned with statistical parameters (standard deviation


and mean value). Taking
into co
nsideration expected electronic noise of HAL25 chip, we are able to find out operation conditions
under which the additional noise contribution becomes negligible.

The effect we are looking at is very small (


0.02*MIP) it makes us to avoid any side distor
tions
capable to hide the effect. Namely settling of the AWG signal must be better 1% within 100ns and the
ALABUF output signal must fit to the full dynamic range of 12
-
bit ADC (
-
1V….+1V). That is why an
attenuators and the second ac
-
coupling used between
the ALABUF and the ADC card. The second ac
-
coupling chain does almost nothing to the signal shape as long as its time constant is very large (

=10uF
1k

=10us).


















Fig. 1.

Experimental set
-
up used for the testing.

100


1k


1k


10uF

10uF

Cin

Cin

The experimental set
-
up.

A
rbitrary

W
aveform

G
enerator


5V range,

Freq=10MHz

O
utput+

Output
-

ALABUF

Gain=5.9

Input+

Input
-

Output+

Output
-

ADC card

PCI
-
DAS 4020/12

[
-
1V…+1V]

Input1

Input2

5


5


Attenuator


20dB

Attenuator


20dB

1.

Calibration (Cin=10uF).


To start with the real measurements an accurate calibration has been carried out first. To do so I
generated two set of number to cover 5V range . Then I loaded the sets into Arbitrary W
aveform
Generator device, which generated stimulus signals for the ac
-
coupled ALABUF chip (see Fig.1, Fig.2).
The ALABUF chip is self
-
biased circuit operating at +1.25V on the input pads. Therefore it cannot be
coupled to an external generator but via capa
citors. For the calibration purpose I used capacitors as large
as 10uF. In this case signal distortion is surely are below level of interest.



















Fig. 2. Calibration. Example

of AGW output signals.


Due to signal attenuator (

0.9) behind of the ALABUF chip its dynamic range seems a little bit
narrower than the actual one (

1.05V) (see Fig.3, Fig.4).

Calibration of ADC Channel #0 (positive),
AWG Channel#1. Scaling factor is (X-2034)/2048
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Sg05P
k
Sg05Pn
k
1
0
k
9
(
)
5
200


Fig. 3. Calibration. Positive output of the ALABUF

chip.

#6

#
4

#2

#6

#4

#5

#2

#1

50mV

100ns

100ns

5V

#100

#5

#3

#1

-
50mV

100ns

100ns

-
5V

AGW
positive
output

AGW
negative
output

#3

Input data
(generated
number).

Scaled output
data (positive
output of the
ALABUF after
analog
-
to
-
digital
conversion).

Uout_pos, V

Uin_pos, V

“zeros”

Calibration of Channel #1(neggative),
AWG Channel #2 . Scaling factor is (X-2079)/2048
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Sg05N
k
Sg05Nn
k
1
0
k
9
(
)
5
200

S1
READPRN
"H:\ASCII_to_WFM\arbascii\signal21"
(
)
1
2.27

0
S4
READPRN
"H:\ASCII_to_WFM\arbascii\sig21n_buf2_10ms"
(
)
2034
(
)
2048


Fig. 4. Calibration. Negative output of the ALABUF chip.


Difference between input and output data lies in range (

3mV) is a result of imperfection of the
AWG generator. This imperfection restricts sensitivity of the exp
erimental set
-
up to the effect of base
line fluctuation. Shift of the values corresponding to even samples (zeros) is 4mV. It occurs due settling
process following the AWG pulse. The greater the pulse is the greater its residual becomes giving
shifting val
ues instead of a fixed level (see Fig 5, Fig.6).


0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
0.01
0.0083
0.0067
0.005
0.0033
0.0017
0
0.0017
0.0033
0.005
0.0067
0.0083
0.01
0.01
0.01
Sg05P
k
Sg05Pn
k
1
0
2.5
0
k
9
(
)
5
200


Fig. 5. Difference between generated data and the ALABUF chip output data. Positive output.






-
Uin_neg, V

Uout_neg, V

Input data
(generated
number).

Scaled output
data (negative
output of the

ALABUF after
analog
-
to
-
digital
conversion).

Uin_pos, V


Uout_pos, V

Imperfection of the AWG

Shift of the“zeros”

End of the ALABUF dynamic
range


0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
0.01
0.0083
0.0067
0.005
0.0033
0.0017
0
0.0017
0.0033
0.005
0.0067
0.0083
0.01
0.01
0.01
Sg05N
k
Sg05Nn
k
1
0
2.5
0
k
9
(
)
5
200



Fig. 6. Difference between generated data

and the ALABUF chip output data. Negative output.



2.

The measurements


MathCad facilities have been used for the data generation. For each pattern there were 768 numbers
generated according to charge left in the strip of the detector. Landau distribution
has been taken into
account as well as 5% occupancy on the strips (see “Study of AC
-
coupling impact on SSD performance
by means of Monte
-
Karlo simulation”). When running at 10MHz frequency the numbers convert into a
set of 100ns pulses (see Fig.7). In tota
l there were 39 patterns coming with 250us gap in between (see
Fig.8).



Fig. 7. An example of the signals to be analyzed.






-
Uin_neg, V


Uout_neg, V

End of the ALABUF dynamic
range

Shift of the“zeros”

Imperfection of the AWG

Genetated signa
l

Signal at the
ALABUF output.

Time, ns




0
1

10
4
2

10
4
3

10
4
4

10
4
5

10
4
6

10
4
7

10
4
8

10
4
9

10
4
0.001
0.0011
0.0012
0.0013
0.0014
0.0015
0.0016
0.0017
0
0.2
0.4
0.6
0.8
1
1
0
T
ms
1.635
10
3

0
ms
100

10
9



Fig. 8. An example of the signals to be analyzed.



Amplitude distribution of the
ALABUF output signals given in Fig.9. A substantial portion of the
small signals is caused by charge division mechanism built
-
in into event generator. The amplitude
distribution becomes Landau distribution like when signals from adjacent channels (strips)
are summed
up. The most significant information on the plot is that MIP is 180 mV therefore expected electronic
noise of HAL25 is


el=180mV 400e/22000e

= 3.3mV




0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
150
300
450
600
536
0
H4a
ka
H45a
ka
1
0
inta
ka


Fig. 9. Amplitude distribution of the ALABUF output signals.

250us

Pattern#1

768st 5%=38events


Pattern#2

Pattern#3

Pattern#4

Pattern#5

768strips 100ns=76.8us

Amplitude distribution of
the ALABUF output signals

Amplitude distribution of the
ALABUF output signals when
signals from

adjacent channels
(strips) are summed up.

Uout_pos, V

MIP=180mV

Entries

.

3. Results of the measurements with Cin=10uF (




10畆






10ms⤮


As it was mentioned for the tests and calibrations there were large capacitors (Cin=10uF) used to
couple the ALABUF chip to AWG generator. It this case difference between generated dat
a and the
ALABUF chip output data caused by the AWG generator imperfection and resolution of the ADC card.
As we can see standard deviation of the difference distribution is


=1.3mV whereas that of expected
electronic noise is

el=3.3mV (see Fig.10). It m
eans that the experimental set
-
up is “sensitive” enough
to observe effect of base line fluctuation we are going to see with smaller coupling capacitors. For the
negative output the resolution is slightly worse (


=1.9mV) hence it is more difficult to obser
ve the
effect after all.


G1
x
(
)
440
exp
x
0.0083
(
)
(
)
2
0.0013
2
1
2


0.03
0.025
0.02
0.015
0.01
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0
200
400
H4
i
G1
x
(
)
N
x
(
)
3

int
i
x

x



Fig. 10. Cin=10uF. Difference between generated data and the ALABUF chip output data. Positive
output.





G2
x
(
)
300
exp
x
0.0125
(
)
(
)
2
0.0019
2
1
2


0.03
0.025
0.02
0.015
0.01
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0
200
400
H5
i
G2
x
(
)
N
x
(
)
2

int
i
x



Fig. 11. Cin=10uF. Difference between generated data and the ALAB
UF chip output data. Negative
output.

Entries

Uout_pos, V

Uout_neg, V

Entries

Expected electronic noise
of HAL25 (

el=3.3mV).

Cin=10uF
. Data
from the positive
output of the
ALAB
UF chip
(

=1.3mV
).

Cin=10uF
. Data
from the negative
output of the
ALABUF chip
(

=1.9mV
).

Expected electronic noise
of HAL25 (

el=3.3mV).

.

4. Results of the measurements with Cin=100nF (




100湆






100us⤮


When capacitors in front of the ALABUF chip are Cin=100nF, the base line fluctuation determine
difference between generated data and the ALABUF
chip output data. Distribution of the differences
becomes wider (


= 2.1mV for the positive output and


=2.4mV for the negative one) (see Fig.12,
Fig.13).



G3
x
(
)
300
exp
x
0.0072
(
)
(
)
2
0.0021
2
1
2


0.03
0.025
0.02
0.015
0.01
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0
200
400
H4
i
H7
i
G3
x
(
)
N
x
(
)
3

int
i
int
i

x

x



Fig. 12. Cin=100nF. Difference between generated data and the ALAB
UF chip output data. Positive
output.




G4
x
(
)
250
exp
x
0.0124
(
)
(
)
2
0.0024
2
1
2


0.03
0.025
0.02
0.015
0.01
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0
200
400
H5
i
H8
i
G4
x
(
)
N
x
(
)
2

int
i
int
i

x

x



Fig. 13. Cin=100nF. Difference between generated data and the ALABUF chip output data. Negative
output.




Entries

Entries

Uout_pos, V

Uout_neg, V

Cin=10uF. Data from the positi
ve
output of the ALABUF chip
(

=1.3mV).

Cin=100nF
. Data
from the positive
output of the
ALABUF chip
(

=2.1mV
).

Expected electronic noise
of HAL25 (

el=3.3mV).

Cin=10uF. Data from the negative
output of the ALABUF chip
(

=1.9mV).

Cin=100nF
. Data from the
negative outp
ut of the
ALABUF chip
(

=2.4mV
).

Expected electronic noise
of HAL25 (

el=3.3mV).

Conclusion.


The measurements carried out with a real set of signals show tha
t additional noise (base line
fluctuations) occurs due to ac
-
coupling in front of the ALABUF chip.

The additional noise becomes “visible” by the experimental set
-
up when coupling capacitor in front
of the ALABUF chip is smaller than 100nF. It is hiding be
hind finite resolution of the experimental set
-
up (

=1.3mV) if the coupling capacitor is much larger than 100nF.

By reconstructing of the amplitude distributions of the ALABUF output signals I determined the
MIP value=180mV and calculated expected electron
ic noise of HAL25
(

el=MIP

400e/22000e=3.3mV).

When the coupling capacitor is 100nF, standard deviation of the additional noise is

=2.1mV (see
Fig.12). Ratio between expected electronic noise and the additional noise is 2.1mV/3.3mV = 0.63. That
is in reas
onable agreement with simulation results 346e/400e = 0.86 (see “Study of AC
-
coupling impact
on SSD performance by means of Monte
-
Karlo simulation”). The discrepancy is most probably caused
by imperfection at the stage of fitting the data with Gaussian func
tions.





































0
5000
1

10
4
1.5

10
4
2

10
4
2.5

10
4
3

10
4
3.5

10
4
4

10
4
4.5

10
4
5

10
4
5.5

10
4
6

10
4
6.5

10
4
7

10
4
7.5

10
4
0
0.2
0.4
0.6
0.8
1
1.147
0
Sig
m
0

Q
j 0
0

0.385
41

76800
0
m
100

Pos
j 0
0

100



100


1k


1k


10uF

10uF

Cin

Cin

Set
-
up.

A
rbitrary

W
aveform

G
enerator


5V range,

Freq=10MHz

Output+

Output
-

ALABUF

Gain=5.9

Input+

Input
-

Output+

Output
-

5


5


Attenuator


20dB

Attenuator


20dB

Event

Generator

A
i

Detector

Simulator

Occupancy


Restored information
over the events

(precise
track

position in space
X
n
i

position in time
T
n
i
)


MathCad software

ADC card

PCI
-
DAS 4020/12

[
-
1V…+1V]

Input2

Input1

The electronics
to test

Real time process
containing pulses coming
out of the front
-
end chip.

Real time process
containing pulses
coming ou
t of the
electronics.

The file

contains description
of pulses coming out of the
fron
-
end chip

Initial information over
the events

(precise track

position in space
X
i

position in time
T
i
)


Generated events

Signals coming out of the front
-
end chip.















= (

2
bl
+

2
el
)
0.5



=
1.2


el


Strip#N+1

Strip#N

Strip#N+5

Strip#N+6

mistake

mistake

Signal before the AC
-
coupling

Signal after the AC
-
coupling


Cin=100nF
.
Difference between generated data and the ALABUF chip
output data.


Entries

Uout_pos, V

Expected electronic noise
of HAL25 (

el=
3.3mV
).

Cin=100nF
. Data
from the positive
output of the
ALABUF chip
(

bl
=2.1mV
).


el=
3.3
mV


bl=
2.1
mV

Report on analogue buffer chip (ALABUF) development in 0.25u CMOS technology
for the ALICE Silicon Strip Detector (SSD).



28 March 2002.

V. Gromov
(vgromov@nikhef.nl), R. Kluit.

ET NIKHEF, Amsterdam.


Abstract.

For the purpose of driving of analog signals from the on
-
detector front
-
end electronics of the ALICE SSD to
the off
-
detector ADC, an analog buffer
chip (ALABUF) has been designed. The design is performed in 0.25


CMOS technology.


Inputs of the design as well as the design goals specification to be met are described along with circuit
optimization procedures and detail chip description.

Results on te
sting of the chips taken from the experimental batch are presented and compared to the
simulations.






Fig.1.
Principal diagram of the on
-
detector electronics of ALICE SSD



Strip#N+1

Strip#N

Strip#N+5

Strip#N+6

mistake

mistake

Signal before the AC
-
couplin
g

Signal after the AC
-
coupling

Time, ns

Fig.2.
The AC
-
coupling effect on the shape of the si
gnals.



Read
-
out rate =10MHz.

N
-
side of the
SSD detector
.

Bias=
0V

P
-
side of the
SSD detector
.

Bias=
55V

ADC

.







Fig.20.
Measurements of linearity and dynamic range of the ALABUF chips.



Fig.2.
Principal diagram of the analog buffer board.



Input differential signal,
InP
-
InN,
mV

Output differential
signal,

OutP
-
OutN


mV

Chip#2, channel 1B

Chip#2, channel 2B

5.6
6 (InP
-
InN)


Fig.24.
Measurements. Differential transient response of the ALABUF chip.

InP, Ch1

InN, Ch2

OutN, Ch3

OutP, Ch4

Settling time 20ns


Talk at the werkbespreking.


ALABUF chip for the ALICE SSD detector.
Design and test.


Vladimir Gromov. ET.

9.04.2003.





Content.


1.

General information.

2.

Function of the ALAB
UF chip.

3.

Principal and schematic diagram.

4.

Main specifications of the chip.

5.

Detector simulator approach to test AC
-
coupling in front of the
chip.

6.

The experimental set
-
up.

7.

The test results discussion.

8.

Conclusion.




Conclusion.


To meet the needs of the rea
d
-
out of the ALICE SSD a new analog
buffer chip ALABUF has been designed in 0.25u CMOS technology.

In order to test the new
-
designed chip we have developed a new
method. The method is based on simulation of the operation of the
detector. The output of this

simulation is a file that describes signal
process coming out the detector. The file is an input for the software
controlled Arbitrary Waveform Generator. The generator produces
“detector
-
like” real time signals for the inputs of the tested chip. The
ADC
card digitizes and saves signal process at the output of the chip
in a file. By taking note of difference between the input and the
output files we analyze signal distortion caused by chip. Thank to
statistical nature of the analysis tiny effects become vi
sible.

I suggest that a many of other applications can make use of this
method.



1
.
1)

ALABUF chip for the ALICE SSD detector. Design and test. Is the
subject of my present talk.

2)

Here are the issues I would like to cover today.

3)

Right after a fe
w general words I am going to tell over the function of
the ALABUF chip in the detector read
-
out.

4)

Then we will have a closer look at the inside of the chip as well as its
performance and its main specifications.

5)

Further I will disclose the substance
of the approach we have developed
to test the chip.

6)

We will examine the experimental set
-
up and will briefly discuss the
test results.

7)

Finally, of course, I will draw some conclusions out this development.


2 .

1)

Since two years ago, VLSI group has
been taking part in project
ALICE.

2)

The group is responsible for development and mass production of two
chips in 0.25u CMOS technology for the Double
-
Sided Silicon strip detector.

3)


Chip I am going to tell is an analog buffer. It is designed for takin
g
analog signals from the front
-
end chips on the both sides of the Silicon strip
detector, amplifying and sending them over 25m twin
-
pair cable to an off
-
detector ADC.

4)

The read
-
out rate will be 10MHZ.

5)

The Front
-
end chips on P
-
side and N
-
side of the

detector will operating
at different potentials.

6)

So as to couple the chips to the rest of the world we had to break DC
-
path and insert capacitors in front of the analog buffer chip.

7)

However the capacitive or AC
-
coupling is never for free, it bring

a lot of
troubles at the same time.
What are they?

8)

Passing over the capacitor each signal leave behind a tail of negative
polarity, which actually is the capacitor discharge.

9)

The tails overlap each other forming a fluctuating substrate for next
com
ing signals.

10)

By making the capacitor larger we make the effect less significant on
the one hand.

11)

But on the other hand then the capacitor will store a huge charge as
long as it holds 55Volts. In the case of accidental break down the will likely
d
estroy all the electronics around here.

12)

Therefore we fall into dilemma. From safety point of view we would
prefer a small capacitor although to minimize signal distortion the capacitor
must be large.

13)

You may often come across the same problem in
your application,
therefore it makes sense to give you a perspective on the approach we have
developed to resolve it.


3 .

1)

First, however, I would like you to have a look at the ALABUF chip
itself.

2)

The analog buffer called ALABUF consists of a feed
-
b
acked operational
amplifier and a pair of analog multiplexers to be able switch inputs to the bus
on any side of the detector.

3)

So as to save power when the chip is out of use a Disenabling option has
been implemented.

4)

Schematic of the amplifier is o
n the plot.

5)

It has a fully differential configuration (differential input and differential
output).

6)

The two
-
stage structure with Miller capacitors guarantees high open loop
gain factor (50dB) and good stability (phase Margin is 83.5 degrees).

7)

R
esistive feedback sets differential gain (5.9).

8)

The common mode feedback takes care that outputs and inputs are kept
at the middle of the power supply range.


4 .

1)

The amplifier stays linear until the output swing does not exceed
1.85V (VDD=2.5V).

2)

The Step response signal smoothly settles to the dc level in the course of
20ns.


5 .

1)


The ALABUF chip houses 2 channels of the analog buffers on area
as large as 2mm by 2mm.


6 .

1)

Main specifications of the analog buffer have been put together in
the Table. All of them comply with the requirements.






7 .

1)
. As soon as the electronics has been designed and preliminary tests
are in line with the simulation you may start to think over a functionality test,
when the ALABUF operates as a link of the

read
-
out chain being set under
real conditions.

2)
. The most common way to provide a real condition tests and to judge
the functionality is to go the real beam of particles. However this way
involves a lot of facilities, time and manpower.

3)
. At the sam
e time we could think of a signal source capable of
simulating processes taking place in the detector.


4)
.
What, in fact
, make the detector on the beam a special signal source,
as long as for the electronics it is just a signal source?

5)
. First we know
that the signals are disorderly spread in the time domain.

6)
. Second the value of the signal varies in a very wide range.

7)
. Third a charge will be shared between two neighbouring strips in a
special manner.


8 .

1)
. If we put all these distribution and
dependences into the MathCAD
software we will generate a file that contains description of pulse coming out
of the “Detector”.

2)
. The Arbitrary waveform generator converts the file into real time
signals and sends them to the AC
-
coupled ALABUF chip.

3)
.
The ADC card does digitizing of the chip output signals.

4)
. We estimate signal distortion by taking difference between signals
coming in and out of the chip.


9 .

1)
. This picture illustrates how the signal patterns look like.

2)
. Each pattern virtually

corresponds to one trigger event.

3)
. It takes 76.8us to read
-
out all the strips.

4)
. After 250us next trigger is coming and the read
-
out starts again.










10 .

1)
. Lets have a look at the test results.

2)
. Here you see statistical distributions
of the mistakes.

3)
. On Y direction are entries, on X direction are the mistake’s value.

4)
. In the first case the coupling capacitor is huge 10uF, the difference
between input and output signals comes from nonideality of the system itself.

5)
. When the

coupling capacitor is smaller 100nF contribution of the base
line fluctuation prevails in the distribution.

6)
. All the mistakes are negative. This is due to the fact that the base line is
sinking a little bit.

7)
. Moreover the distribution has a nonzero

width that means that the base
line fluctuates.

8)
. We may consider the base line fluctuation as additional noise .

9)
. So as to evaluate it we fit it this distribution with a Gaussian curve.

10)
. Standard deviation or root mean squire value is a figur
e to compare
the base line fluctuation to the expected electronic noise that is always present
in here.

11)
. According to the general rule to sum independent random processes
the total noise goes to 3.9mV.

12)
. It yield 20% addition the electronic noise.

To make this addition
smaller we must choose for the larger capacitor
.




11 .

Read CONCLUSION.


.



To figure out what the real conditions are we must recall that signals coming out of the detector are of random
amplitude and expected to be a random pr
ocess in the time domain.

As an engineer I design electronics for particle detectors. I always want to make sure that the electronics is the best to fu
lfill functions it is designed
for.

What are these functions? In general, the electronics together with

the particle detector constitute a device by means of which we are able to obtain
some information over the particles. A Tracking system reconstructs tracks of the particles and a Calorimeter measures energy
. In reality the information is
wrapped in pulse
s coming out of the detector. The function of the electronics is to process the incoming pulses in order to unwrap the inform
ation over the
particles in the best way.

How can I judge the functionality of the electronics?

The most common way to provide a re
al condition tests and to judge the
functionality is to go the real beam of particles. However this way involves a lot of facilities, time and manpower. At the
same time we could think of a signal source capable of simulating processes taking place in the
detector. Having an Event
Generator inside such a source could generate an array with initial information over the events (precise track position for
instance) and a file, which contains description of pulses coming out of the front
-
end chip. This file is
an input for the
Arbitrary Waveform Generator to produce a real time process to be a stimulus for the AC
-
coupled ALABUF chip. I could
use signals at the output to restore the initial information. Difference between initial information and restored informat
ion
numerically characterizes distortions caused by the tested electronics. I may consider the difference as a figure of merit wh
en
tuning the electronics in order to reach the best functionality
.





This approach has been developed to test electronics fo
r the ALICE Silicon Strip Detector. A software signal source
plays the key role in the set
-
up. It is made on the basis of MathCad software and involves:

1)

random distribution of the events in time according to Poisson statistics

2)

Landau statistics of the char
ge deposition in each event

3)

mechanism of charge sharing between adjacent strips of the detector.




Besides the software signal source the experimental set
-
up includes an Arbitrary waveform generator (12 bit in range
50mV…5V, clock frequency up to 250MHz),

an ADC card (12 bit in range

1V…+1V). The electronics to test is an analog
buffer chip ALABUF and AC

coupling network in front of it.

What in particular we are going to look at. It is to AC or capacitive coupling that each signal is followed by a long t
ail
of negative polarity, which in fact is discharge the capacitor. The tails pile up each other thus forming a base line fluctua
tion.
The fluctuation is actually noise that spoils performance of the detector. For safety and space reasons it is preferable
to keep
values of the capacitors as low as possible. On the other hand the smaller the capacitor is the more substantial the base lin
e
fluctuation noise becomes. Our goal is to determine a minimum value where the base line fluctuation noise gives a not yet

noticeable addition to the intrinsic electronic noise.


Results of the measurements are given here. Difference between primary (generated) data and data restored after
passing through the electronics to test is a histogram with a fitting curve on top of
it. As for the fitting function it as a normal
Gaussian distribution. The standard deviation parameters of the distribution are the values to compare with those of expected

electronic noise.

When the capacitor is enormously big (10uF), difference between
primary and restored data is not caused by the tested
items but the finite resolution of the set
-
up. This is the ideal case. If the capacitors are becoming smaller the base line noise is
approaching the expected electronic noise. However at the value of 10
0nF the standard deviation of the expected electronic
noise is still 1.5 times as large that for the base line fluctuation noise.